ES2858306T3 - Method for determining HLA status by sequencing the immune repertoire - Google Patents

Abstract

A method for predicting the status of a subject's human leukocyte antigen (HLA) allele, comprising: (a) determining an immune receptor profile of T-cell receptor (TCR) rearranged DNA sequences unique to each of a plurality of subjects, each subject having a known HLA allele status; (b) categorizing the plurality of subjects based on (i) said subject's known HLA allele status and (ii) a presence or absence in the subject's immune receptor profile of a trait comprising a single TCR rearranged DNA sequence ; (c) determining a statistical score for the association between a set of traits and an HLA positive allele status based on (b); (d) training a machine learning model using said set of traits to define a set of classifiers for each HLA allele state; (e) introducing one or more rearranged DNA sequences from the unique TCR of a subject with an unknown HLA allele status into said machine learning model to identify one or more traits that match the set of classifiers; and (f) predicting the status of an HLA allele of said subject based on said one or more paired traits.

Description

DESCRIPTION

Method for determining HLA status by sequencing the immune repertoire

Background of the invention

The cellular adaptive immune system conveys broad protection against infection by pathogens through the development of a vast and highly diverse repertoire of T-cell receptor (TCR) genes, which encode cell-surface T-cell receptors with randomized antigenic specificity. The ability of a subject's adaptive immune system to adequately address an early infection is based on the activation of an appropriate antigen-specific T-cell receptor (TCR). The TCR-antigen interaction is mediated by the presentation of foreign peptides on the cell surface by cells infected with pathogens in the environment of major histocompatibility complex (MHC) proteins. Specifically, CD8 + T cells recognize the antigen in the environment of MHC class I proteins. Since MHC class I proteins are encoded by human leukocyte antigen (HLA) loci A, B, and C, which are highly polymorphic, the antigenic specificity of a TCR is modulated between individuals by the HLA environment.

When an antigen has been found, activated T lymphocytes proliferate by clonal expansion and reside in the memory T lymphocyte space for many years as a clonal population of cells (clones) with rearranged TCR genes identical by descent (Arstila TP, et al. . A direct estimate of the human alphabeta T cell receptor diversity. Science 286: 958-961, 1999). Protection against future exposure to disease-causing pathogens is conferred by the ability of activated T cells to form durable memory responses.

Most of the diversity of TCRs resides in the p chain of the alpha / beta heterodimer of TCRs. Each T-cell clone is encoded by a single TCRp allele that has been randomly rearranged from the germline TCRB locus to form a mature TCRp gene. Immense diversity is generated by combining the variable (V), diversity (D), and binding (J) TCRp gene segments, which collectively encode the CDR3 region, the primary region of the TCRB locus to determine antigen specificity. The template-independent deletion and insertion of nucleotides during rearrangement at the Vp-Dp and Dp-Jp junctions adds to the potential diversity of receptors that can be encoded (Cabaniols JP, et al. Most alpha / beta T cell receptor diversity is due to terminal deoxynucleotidyl transferase. J Exp Med 194: 1385-1390, 2001). Typically, at any given time, a healthy adult expresses approximately 10 million unique TCRp chains in their 1012 circulating T cells (Robins HS, et al. (2009) Comprehensive assessment of T-cell receptor beta-chain diversity in alphabeta T cells. Blood 114: 4099-4107). However, observing the same TCRp chain independently in two individuals is thousands of times more common than would be expected if all rearrangements were equally likely (Robins HS, et al. (2010) Overlap and effective size of the human CD8 + T cell repertoire receptor. Science Translational Medicine 2: 47ra64). It is expected that many TCRp sequences (especially those with few or no insertions) will be present in the unexposed repertoires of most individuals and that these TCRp sequences will reliably proliferate upon exposure to their associated antigen ( V. Venturi, et al., The molecular basis for public T-cell responses? Nature reviews. Immunology 8, 231-238 (2008); published online EpubMar (10.1038 / nri2260)). This overrepresentation of specific TCRp sequence rearrangements in the T-cell repertoire without prior exposure forms the basis for public T-cell responses.

Public T cell responses occur when T cells bearing identical T cell receptors (TCRs) are observed to dominate the response to the same antigenic epitope in multiple individuals. Many pathogenic antigens are known to induce such a public T cell response, in which a pathogenic antigen is targeted by the same T cell receptor sequence (and found to be immunodominant) in multiple individuals with specific HLA isotypes. H. Li et al., Determinants of public T cell responses. Cell research 22, 33 42 (2012); published online EpubJan (10.1038 / cr.2012.1); H. Li, et al., Recombinatorial biases and convergent recombination determine inter-individual TCRbeta sharing in murine thymocytes. Journal of immunology (Baltimore, Md .: 1950) 189, 2404-2413 (2012); published online EpubSep 1 (102087). In other words, public T-cell responses are observed when the potential high avidity TCRp chain space that could bind to a particular antigen-MHC complex includes one or more TCRp chains that also have a high probability of existing in the repertoire. no prior exposure at any given time. Therefore, sequences associated with a public T-cell response will only be present intermittently in space without prior exposure of subjects who have not been exposed to a particular antigen; however, such clones should appear constantly in the repertoire of T lymphocytes of subjects who have been exposed to the antigen, having undergone clonal expansion after stimulation of the antigen.

Previous work on public T-cell responses has identified individual examples of public T-cell responses to disease (including CMV, EBV, influenza, multiple sclerosis, and other neoplastic diseases and autoimmune conditions), (Venturi V, et al. (2008) J Immunol 181: 7853-7862). These studies were limited by the depth of sequencing and the size of the research cohorts. Typically, these public T-cell responses were studied in the environment of individual antigens in a unique HLA environment, typically using MHC-antigen tetramers to purify antigen-specific T lymphocytes. However, due to technical limitations, to date, only a relatively small number of public T cell responses have been identified. Not surprisingly, these results have been limited to high frequency and readily observable public T cell responses that they dominate the immune response to their target antigens. Therefore, these public responses represent only TCRp sequences that are common to the T-cell space without prior exposure of almost all adult subjects (i.e., available for antigen response), and they miss both rare T-cell rearrangements that are present only in some antigen-experienced patients due to the rarity between previously exposed T-cell sequences and the responses of T-cells that are not immunodominant and therefore never comprise a sufficient fraction of the total T-cell repertoire to be reliably observed with limited sequencing depth.

Furthermore, the TCR-antigen interaction is mediated by the presentation of foreign peptides on the cell surface by cells infected with pathogens in the environment of major histocompatibility complex (MHC) class I proteins. are encoded by the human leukocyte antigen (HLA) loci A, B and C, which are highly polymorphic, the antigenic specificity of a TCR is modulated between individuals by the HLA environment.

The binding of T cell receptors to antigens is mediated by MHC proteins, which present the antigen on the surface of cells. The MHC are encoded in humans by HLA loci, which are highly polymorphic. This polymorphism results in heterogeneous T cell responses to the same antigen in all individuals and a positive and negative differential selection of specific T cell receptor sequences during thymic training. The determination of an individual's HLA alleles (HLA type) has several clinical applications. An example of a clinical application for HLA typing is testing an individual's suitability as a bone marrow transplant donor. HLA typing is described in Spellmann et al., Biol Blood Marrow Transplant. 2008; 14 (Suppl 9): 37-44; Bidwell, Immunol Today. 1994; 15 (7): 303-307; and WO2011017151.

There is a need for improved methods to diagnose and / or predict a person's status for an infectious disease, such as CMV, EBV, HPV, smallpox, and others, with greater sensitivity and precision. Diagnostic methods are needed that take advantage of information about an individual's T-cell receptor sequence profile, including the presence of public T-cell clones, and assess the individual's infectious disease status based on the sequence profile. There is also a need to infer the HLA type of an individual based on the individual's T cell receptor sequence profile. The present invention satisfies these needs and provides additional related advantages.

Summary of the Invention

The invention provides a method for predicting the status of a human leukocyte antigen (HLA) allele of a subject, comprising:

(a) determining an immune receptor profile of unique T-cell receptor (TCR) rearranged DNA sequences for each of a plurality of subjects, each subject having a known HLA allele status; (b) categorizing the plurality of subjects based on (i) said subject's known HLA allele status and (ii) a presence or absence in the subject's immune receptor profile of a trait comprising a single TCR rearranged DNA sequence ;

(c) determining a statistical score for the association between a set of traits and an HLA positive allele status based on (b);

(d) training a machine learning model using said set of traits to define a set of classifiers for each HLA allele state;

(e) introducing one or more rearranged DNA sequences from the unique TCR of a subject with unknown HLA allele status into said machine learning model to identify one or more traits that match the set of classifiers; and

(f) predicting the status of an HLA allele of said subject based on said one or more paired traits. The present disclosure is based, in part, on methods for determining or predicting the presence or absence of one or more infectious disease agents in a subject with known or unknown infection status by immune quantification techniques combined with mathematical modeling.

The present disclosure provides a method of predicting the presence or absence of an infection in a subject with unknown infection status. In one embodiment, genomic DNA from a sample comprising T lymphocytes obtained from the subject is subjected to high-throughput amplification and sequencing to determine a T lymphocyte receptor (TCR) profile comprising amino acid sequences of the complementarity determining region 3 (CDR3) of unique TCRs. In a particular case, the TCR profile is then compared to a database of previously identified diagnostic public T-cell receptor sequences known to be statistically significantly associated with infection. In one case, Next, a first score is generated for the subject by determining the proportion of unique TCR sequences in the subject's profile that match the public TCR sequences in the database. In this case, the first score is entered into an algorithm that compares the subject's first score with the known infection states of a plurality of subjects with known infection status. In another case, then, an estimated probability of infection status for the subject is determined as a result of the algorithm. On certain occasions, the method further comprises an initial step of obtaining a sample comprising T lymphocytes from the subject.

The present disclosure further provides a method for predicting the presence or absence of one or more viral infections in a subject with unknown infection status. In another case, the method comprises determining a profile of unique TCR sequences from a sample obtained from the subject (eg, a sample comprising T lymphocytes) and inputting these unique TCR sequences into one or more algorithms. In one instance, the one or more algorithms are generated by determining at least 100,000 unique TCR sequences from each of a plurality of subjects of known infection status for each of the one or more infections. On certain occasions, the method further comprises an initial step of obtaining a sample from the subject. On certain occasions, unique TCR sequences are statistically identified that correlate with the presence or absence of each of the one or more infections, thus generating a score that predicts the presence or absence of each of the one or more infections. In these cases, the scores are then entered into a logistic regression model trained in each of the plurality of subjects of known infection status for each of the one or more infections, the result of which is the prediction of whether the subject is positive. or negative for each of the one or more infections.

In some cases, one or more infections are a cytomegalovirus (CMV), an Epstein-Barr virus (EBV), a herpes simplex virus (HSV), or a smallpox virus.

Described herein is a method for predicting the presence or absence of a CMV infection in a subject with unknown infection status. In one instance, genomic DNA from a sample comprising T lymphocytes obtained from a subject is subjected to high throughput amplification and sequencing to determine a TCR profile comprising unique CDR3 amino acid sequences. On certain occasions, the method further comprises an initial step of obtaining a sample comprising T lymphocytes from the subject. In other cases, the TCR profile is then compared to a database of previously identified diagnostic public T-cell receptor sequences known to be statistically significantly associated with CMV infection. On certain occasions, a CMV loading score is then generated for the subject by determining the proportion of unique TCR sequences in the subject's profile that match the public TCR sequences in the database. In other cases, the calculated CMV burden score is entered into a logistic regression model that compares CMV burden and CMV infection status of a plurality of subjects with known CMV infection status. In one case, an estimated probability of the CMV infection status for the subject is then determined as a result of the logistic regression model.

On certain occasions, the public T cell sequence database is determined to be statistically associated by obtaining unique TCR sequences from a group of subjects with CMV infection and a group of subjects without CMV infection. In one case, each unique TCR sequence is subjected to a one-tailed Fisher's exact test based on the presence or absence of each TCR sequence in subjects with known CMV infection status. In one case, the null hypothesis is that each TCR sequence is no more common in subjects with a CMV infection than in subjects without a CMV infection. In one case, a theoretical threshold of p-value is established, and the false discovery rate (TDF) is controlled by permutation of the CMV infection status in each subject to generate an empirical null distribution of p-values. In a particular case, a database is then generated comprising T-lymphocyte receptor sequences that are statistically significantly shared in subjects with CMV infection.

On certain occasions, the theoretical threshold of p-value is less than or equal to 1.0 * 10-4.

On certain occasions, statistically identified unique TCR sequences that correlate with the presence or absence of CMV infection comprise one or more of SEQ ID NO: 1 to 142. On certain occasions, statistically identified unique TCR sequences that are correlate with the presence or absence of CMV infection comprise SEQ ID NO: 1 to 142.

In some embodiments, the step of determining the at least 100,000 unique TCRp sequences from the sample includes amplifying the rearranged nucleic acids encoding the CDR3 region of TCRp in a multiplexed PCR reaction with a mixture of gene segment-specific forward primers. TCR Vp and specific reverse primers of the Jp gene segments. In some embodiments, the amplified nucleic acid reads are sequenced and the sequence reads are processed to eliminate errors in the primary sequence of each read and compress the data. In particular embodiments, a nearest neighbor algorithm is applied to collapse the data into unique sequences by fusing closely related sequences to eliminate both PCR and sequencing errors.

The method also includes determining an HLA association for each unique TCR sequence in the TCR profile. In some embodiments, the unique TCR sequence is associated with an HLA-A and / or HLA-B allele.

The method of the invention comprises steps of predicting the status of a subject's human leukocyte antigen (HLA) allele, comprising (a) determining an immune receptor profile of unique T-cell receptor (TCR) rearranged DNA sequences. for each of a plurality of subjects, each subject having a known HLA allele status; (b) categorizing the plurality of subjects based on (i) said subject's known HLA allele status and (ii) a presence or absence in the subject's immune receptor profile of a trait comprising a single TCR rearranged DNA sequence ; (c) determining a statistical score for the association between a set of traits and an HLA positive allele status based on (b); (d) training a machine learning model using said set of traits to define a set of classifiers for each HLA allele state; (e) introducing one or more rearranged DNA sequences from the unique TCR of a subject with unknown HLA allele status into said machine learning model to identify one or more traits that match the set of classifiers; and (f) predicting the HLA allele status of said subject based on said one or more matching traits.

The method comprises determining an immune receptor profile by determining the total number of unique TCR sequences and the frequency of each unique TCR sequence. The method also comprises determining a statistical score which comprises determining a p-value using a two-tailed Fisher's exact test.

In some embodiments, the method includes determining a p-cutoff value to identify a set of traits that are significantly associated with an HLA allele status. In another embodiment, the method includes determining a false discovery rate (TDF) of the association of a trait with an HLA allele status. In other embodiments, the method includes determining a series of false positive associations between said trait and said HLA allele status.

In some embodiments, the method includes training a machine learning model by training a logistic regression model using said set of identified characteristics and said known HLA allele states of each subject. In one embodiment, the method includes training a machine learning model that comprises performing a cross-validation method of leaving one out. In another embodiment, the method comprises performing said cross-validation method of leaving one out for multiple rounds. In another embodiment, the prediction is at least 80% accurate or at least 90% accurate.

In certain embodiments, the TCR rearranged DNA sequence is a TCRA, TCRB, TCRG, or TCRD rearranged DNA sequence. In other embodiments, the HLA allele is an HLA-A2 allele or an HLA-24 allele.

Brief description of the drawings

A better understanding of the novel features of the invention and the advantages of the present invention will be obtained by referring to the following description and the accompanying drawings of which:

Figure (FIG.) 1 depicts eleven different CDR3p VDJ recombination events, while still producing the same nucleotide sequence and ultimately the same amino acid sequence of CDR3p (SEQ ID NO: 149). The nucleotide sequences on the left side of FIG. 1 are, in order from top to bottom, SEQ ID NO: 143, SEQ ID NO: 144, SEQ ID NO: 144, SEQ ID NO: 144, SEQ ID NO: 144, SEQ ID NO: 144, SEQ ID NO: 145, SEQ ID NO: 146, SEQ ID NO: 146, SEQ ID NO: 147, SEQ ID NO: 148.

FIGS. 2A, 2B, and 2C provide an overview of the method for predicting an individual's disease state. FIG. 2 shows a set of data samples from peripheral blood of 640 healthy subjects (287 and 353 CMV- CMV +), which were analyzed by immune receptors profiles TCR high performance. In FIG. 2B, unique TCRp sequences were identified that were present in significantly more CMV + subjects than CMV- subjects, controlling for the false determination rate (TDF) by permutation of CMV status. The presence of these CMV-associated TCRp sequences was used to construct a classification model. The top panel of FIG. 2B represents CASSLIGVSSYNEQFF (SEQ ID NO: 12). In FIG. 2C, the classification model was tested using extensive cross-validation of leaving one out, in which a sample was kept out of the calculations and the process was repeated from the beginning. The resulting classification model was used to predict the c Mv status of the excluded subject.

FIG. 3A shows a graphical representation of a subject data set, where 640 subjects were phenotyped specifically for HLA type and CMB status to be eligible as HCT donors, and were separated into 287 CMV + donors and 353 CMV- donors. . There were approximately the same number of seropositive and seronegative samples for the investigation of public T cell responses.

FIG. 3B shows the demographic characteristics of the subjects included in this study, classified by CMV status.

FIG. 4 shows a test of association to CMV of the TCRs, where the amino acid sequence of the CASSLIGVSSYNEQFF TCR (SEQ ID NO: 12) was selected and identified in 27 of the CMV + subjects, and only 2 of the CMV- subjects, with a p-value of 2.8E'08.

FIG. 5 represents amino acid sequences of the illustrative TCRs identified in CMV + subjects with a threshold p-value chosen at 10 "4 In order from top to bottom, these are SEQ ID NO: 1-5. It can be seen that at p -lower value, a greater number of TCRp sequences associated with CMB can be identified. FIG. 6 represents the CMV load, which is represented as the proportion of unique TCRs in each subject that are significantly associated with ^c M ^v versus to those TCRs that are not significantly associated with CMV.

FIG. 7 describes the cross-validation method, which leaves out one of the initial 640 samples, then the CMV-specific TCRp sequence database and associated statistics are retrained in an unsupervised manner to eliminate bias, and The CMV serological status of the excluded sample is classified. FIG. 7 represents CASSLIGVSSYNEQFF (SEQ ID NO: 12).

FIG. 8 describes the results of the cross-validation method shown for all subjects and cross-validated subjects. Data are presented as the area under the receiver operating characteristic (ABCOR) on the y-axis versus the p-value on the x-axis. The results of the cross-validation method are displayed as a plot of the false positive rate versus the true positive rate. FIG. 8 (upper graph) shows performance data for the classification of all datasets and cross - validation (CV) for each threshold p-value, measured as the area under the curve ^c O ^r (Abcor). The number on each set of data points corresponds to the number of TCRp associated with CMV identified at that p-value threshold, and the rectangle indicates the data set selected for subsequent analysis (p-value = 10-4) . FIG. 8 (bottom graph) also shows an estimated false discovery rate (TDF) for each threshold p-value used in the identification of TCRp sequences significantly associated with CMV, using permutations of CMV status. The best performance is observed with a p-value of 10-4, which corresponds to an estimated TDF of ~ 20%, resulting in the identification of a set of 142 TCR-p sequences that were significantly associated with the positive state. CMV (listed in Table 1).

FIG. 9 shows the COR curves for the total and cross-validation data sets.

FIGS. 10A and 10B show HLA restriction of CMV-associated TCRp sequences. FIG. 10A shows the distribution of HLA-A alleles in this cohort. FIG. 10B shows the distribution of HLA-B alleles in this cohort. FIGS. 10A and 10B represent the following sequences: CASSLAPGATNEKLFF (SEQ ID NO: 8), CASSLIGVSSYNEQFF (SEQ ID NO: 12), CASSPSRNTEAFF (SEQ ID NO: 73), CASSLQAGANEQFF (SEQ ID NO: 119) and CASASANYGYTF (SEQ ID NO: 118).

FIGS. 11A and 11B show the incidence of previously documented CMV-reactive TCRp sequences in this cohort. FIG. 11A shows the incidence of each sequence of PCRRs in the cohort of 640 subjects represented graphically along the horizontal axis by decreasing the total incidence, with the incidence in CMV + subjects above the horizontal and the incidence in CMV- subjects. below horizontal. FIG. 11B shows an incidence histogram of these sequences of the TCRp in the cohort of 640 subjects plotted for each group of sequences.

FIG. 12 shows the concordance of the TCRB sequences in the cohort compared to those in the literature. FIG. 12 represents the following sequences: CASSLAPGATNEKLFF (SEQ ID NO: 8), CASSLIGVSSYNEQFF (SEQ ID NO: 12), CASSPSRNTEAFF (SEQ ID NO: 73), CASSLQAGANEQFF (SEQ ID NO: 119) and CASASANYGYTF (SEQ ID NO: 118) .

FIG. 13 represents a general description of an embodiment of the method. FIG. 13 represents CSARDRGIGNTIYF (SEQ ID NO: 152).

FIG. 14 depicts the selection of traits using a two-tailed Fisher's exact test to determine the statistical significance of the association between a trait (unique TCRB sequence) and HLA allele status (HLA-A2 + or HLA-A2-), of according to an embodiment of the invention. An illustrative listing of characteristics (unique TCRB sequences) and the number of subjects that have a particular characteristic and whether the subject is positive or negative for the HLA-A2 allele is shown. FIG. 14 represents, in order from top to bottom, SEQ ID NO: 150-161.

FIG. 15 depicts the steps for selection of traits, according to one embodiment of the invention. A p-value is selected as a cut-off point to identify a set of "TCR traits" from the complete list of possible TCR sequences. Defining a threshold p-value and permutation of allele status between individuals provides an estimate of the false discovery rate. This is done for each HLA allele, resulting in a set of TCRp sequences associated with the alleles for each HLA allele. A p-cutoff value of p <10.4 and a TDF of 0.1 were used to identify 288 TCRB sequences that are positively associated with HLA-A2. For each of the TCRB sequences associated with the alleles, the frequency of the sequence is also determined in each subject. FIG. 15 represents, in order from left to right, SEQ ID NO: 162 and 163.

FIG. 16 depicts a machine learning process for fitting a logistic regression model, in accordance with one embodiment of the invention.

FIG. 17 represents a comprehensive cross-validation of leaving one out, in accordance with one embodiment of the invention.

FIG. 18 shows the results of the cross-validation experiment and illustrates the precision of the method.

Detailed description of the invention

This document describes methods for predicting and diagnosing infectious diseases in a subject that they are disease specific (specific for each particular disease state), they have universal support (they do not require different processes / reagents for each disease state) and they are multiplexed (they are capable of analyzing multiple disease states simultaneously).

In one case, the method includes steps for high-throughput immunosequencing of rearranged TCR genes in healthy subjects with known CMV status. Using the results of immunosequencing, the method includes searching for TCRp sequences that are present in multiple subjects and identifying a set of TCRp sequences that are significantly associated with CMV positive status. The method also includes calculating a p-value for the association of each TCRp sequence with CMV status using Fisher's exact test, controlling the false discovery rate (TDF) by permutation of CMV status, and identifying a listing of TCRp sequences associated with CMV (for a given TDF and p-value). A CMV score for each subject is calculated as the proportion of all TCRp sequences from that subject that are represented in the catalog of CMV-associated TCRp sequences. The CMV score is used to distinguish between CMV + and CMV- subjects.

Infectious agents include pathogens, viruses, bacteria, parasites, and / or microorganisms. Viruses sometimes include, but are not limited to, members of the herpes virus family (such as herpes simplex viruses 1 and 2, varicella-zoster virus, EBV (Epstein-Barr virus), human cytomegalovirus (CMV), human herpes virus 6, human herpes virus 7 and herpes virus associated with Kaposi sarcoma), hepatitis B virus (HBV), hepatitis C virus (HCV), human immunodeficiency virus (HIV ) I and II, influenza virus A, influenza virus B, respiratory syncytial virus (RSV) A and B, and human metapneumovirus (MPV). Other examples include human T-cell lymphocytotrophic virus, human papillomavirus, orthomyxovirus, paramyxovirus, adenovirus, coronavirus, rhabdovirus, polio virus, togavirus, buniavirus, arenavirus, rubella virus, reovirus, norovirus, human metapneumovirus (MPV) , West Nile virus, yellow fever virus, rabies virus, rhinovirus, Rift Valley fever virus, Marburg virus, mumps virus, measles virus, human papillomavirus (HPV), virus Ebola, Colorado tick fever virus (CTFV) and / or rhinovirus.

Other infectious organisms include Escherichia coli, Salmonella, Shigella, Campylobacter, Klebsiella, Pseudomonas, Listeria monocytogenes, Mycobacterium tuberculosis, Mycobacterium avium-intracellulare, Yersinia, Francisella, Pasteurella, Brucella, Clostridia, Bordetella pertussis, Stacousterophy pneumonia, Streptocus acoureus, Streptococcus group B haemolytic, Corynebacteria, Legionella, Mycoplasma, Ureaplasma, Chlamydia, Clostridium difficile, Gardnerella, Trichomonas vaginalis, Neisseria gonorrhea, Neisseria meningitides, Hemophilus influenza, Enterococcus faecalis, Proteus vulgardor, Tremoris pallabilis, Proteicobacteria palmaris, Proteicobacteria palmaris, Proteicobacteria palladium Borrelia recurrentis, rickettsial pathogens, Nocardia, Acitnomycetes and / or Acinetobacter.

In still other cases, fungal infectious agents include, but are not limited to, Cryptococcus neoformans, Blastomyces dermatitidis, Histoplasma capsulatum, Coccidioides immitis, Paracoccicioides brasiliensis, Candida albicans, Aspergillus fumigautus, Phycomomyycetes (Rhizopuschen.

In more cases, parasitic agents include, but are not limited to, Plasmodium falciparum, Plasmodium malaria, Plasmodium vivax, Plasmodium ovale, Onchoverva volvulus, Leishmania, Trypanosoma spp., Schistosoma spp., Entamoeba histolytica, Cryptosporidum, Giardia spp., Trichimonas spp. , Balatidium coli, Wuchereria bancrofti, Toxoplasma spp., Enterobius vermicularis, Ascaris lumbricoides, Trichuris trichiura, Dracunculus medinesis, trematodes, Diphyllobothrium latum, Taenia spp., Pneumocystis carinitori and / or Necanis america.

The major histocompatibility complex (MHC) is a set of cell surface molecules encoded by a large family of genes that controls an important part of the immune system in all vertebrates. The primary function of major histocompatibility complexes is to bind peptide fragments from pathogens and display them on the cell surface for recognition by the appropriate T lymphocytes. The major histocompatibility complex (MHC) contains two types of MHC polymorphic genes, class I and class II genes, which encode two groups of structurally distinct but homologous proteins, and other non-polymorphic genes whose products are involved in presentation. antigens.

Human MHC is called human leukocyte antigen (HLA). HLA proteins are encoded by MHC genes. HLA class I antigens include HLA-A, HlA-B, and HLA-C. HLA class II antigens include HLA-DR, HLA-DQ, HLA-DP, HLA-DM, HLA-DOA, HLA-DOB. The HLAs corresponding to MHC class I (A, B and C) present peptides from the interior of the cell. HLAs corresponding to MHC class II (DP, DM, DOA, DOB, Dq, and DR) present antigens from outside the cell to T lymphocytes. MHC molecules mediate the binding of a given T lymphocyte receptor to an antigen. given, therefore, the MHC polymorphism between individuals modulates the response of T lymphocytes to a given antigen. The clinical applications of HLA typing include vaccine trials, disease associations, adverse drug reactions, platelet transfusion, and organ and stem cell transplantation.

There are numerous alleles at each HLA gene locus (A1, A2, A3, etc.). Each person inherits a complete set of HLA alleles (haplotype) from each parent, and this combination of encoded proteins constitutes a person's HLA type (for example, different antigens from A23, A31, B7, b44, C7, C8, DR4 , DR7, DQ2, DQ7, Dp2, DP3). There are more than 50,000 different types of HLA known.

As used herein, adaptive immune receptor (AIR) refers to an immune cell receptor, for example, a T-lymphocyte receptor (TCR) or an immunoglobulin receptor (Ig). found in mammalian cells. In certain embodiments, the adaptive immune receptor is encoded by a gene or gene segment from TCRB, TCRG, TCRA, TCRD, IGH, IGK, and IGL.

The term "primer", as used herein, refers to an oligonucleotide sequence capable of acting as a starting point for DNA synthesis under suitable conditions. Such conditions include those in which the synthesis of a primer extension product complementary to a nucleic acid strand is induced in the presence of four different nucleoside triphosphates and an extension agent (eg, a DNA polymerase or reverse transcriptase) in a appropriate buffer and at a suitable temperature. As used herein, the term "gene" refers to the segment of DNA involved in the production of a polypeptide chain, such as all or a portion of a TCR or Ig polypeptide (eg, a polypeptide containing CDR3); includes regions preceding and following the "leader and tail" coding region, as well as intervening sequences (introns) between individual coding segments (exons), and may also include regulatory elements (eg, promoters, enhancers, binding sites repressor and the like), and may also include recombination signal sequences (RSS), as described herein.

The nucleic acids of the present embodiments, also referred to herein as polynucleotides, and including oligonucleotides, can be in the form of RNA or in the form of DNA, including cDNA, genomic DNA, and synthetic DNA. DNA can be double-stranded or single-stranded, and if it is single-stranded it can be the coding strand or the non-coding strand (antisense). A coding sequence encoding a TCR or an immunoglobulin or a region thereof (eg, a V region, a D segment, a J region, a C region, etc.) for use in accordance with the present embodiments may be identical to the coding sequence known in the art for any given TCR or immunoglobulin gene region or polypeptide domains (eg, V region domains, CDR3 domains, etc.), or may be a coding sequence different, which, as a result of the redundancy or degeneracy of the genetic code, encodes the same region or polypeptide of the TCR or of the immunoglobulin.

Unless specific definitions are provided, the nomenclature used in relation to laboratory procedures and techniques of, molecular biology, analytical chemistry, synthetic organic chemistry, and medicinal and pharmaceutical chemistry described herein are those well known and commonly used in the technique. Standard techniques can be used for recombinant technology, molecular biology, microbiology, chemical synthesis, chemical analysis, pharmaceutical preparation, formulation, and administration and treatment of patients.

Unless the context requires otherwise, throughout this specification and claims, the word "comprise" and variations thereof, such as, "comprises" and "comprising" are to be construed in an open, inclusive sense. , that is, as "which includes, but is not limited to". By "consisting of" is meant to include, and is normally limited to, what follows in the term "consisting of". By "consisting essentially of" is meant to include any items listed after the expression, and is limited to other items that do not interfere with or contribute to the activity or action specified in the disclosure of the listed items. Therefore, the expression "consisting essentially of" indicates that the listed items are required or required, but that other items are not required and may or may not be present depending on whether or not they affect the activity or action of the listed items. .

In this specification and in the appended claims, the singular forms "a", "an", "the" and "the" include plural references, unless the content clearly indicates otherwise.

Reference throughout the present specification to "a single embodiment" or "one embodiment" or "one aspect" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the embodiment. present invention. Thus, the occurrences of the terms "in one embodiment" or "in one embodiment" in various places throughout the present specification do not necessarily refer to the same embodiment. Furthermore, the particular features, structures, or features can be combined in any suitable way in one or more embodiments.

Methods of the invention

Cells

A sample can be obtained that contains lymphoid nucleic acid molecules (genomic DNA, cDNA, or, as alternative, messenger RNA) from a subject. The subject is a mammalian subject, such as a human.

Lymphocytes (B lymphocytes and / or T lymphocytes) can be obtained from a biological sample, such as from a variety of biological fluid and tissue samples. These include but are not limited to the bone marrow, thymus, lymph glands, lymph nodes, peripheral tissues, and blood or solid tissue samples. A sample can be taken from any peripheral tissue.

use in the methods described herein. Peripheral blood mononuclear cells (PBMC) are isolated by techniques known to those of skill in the art, for example, by Ficoll-Hypaque® density gradient separation. In certain embodiments, full P ^bm C are used for analysis.

Nucleic acid extraction

Total genomic DNA can be extracted from cells by methods known to those of skill in the art. Examples include the use of the QIAamp® DNA Mini Blood Kit (QIAGEN®) or a Qiagen DNeasy Blood Collection Kit (Qiagen, Gaithersburg, MD, USA). The approximate mass of a single haploid genome is 3 pg. Preferably, at least 100,000 to 200,000 cells are used for diversity analysis, ie, about 0.6 to 1.2 pg of diploid T-lymphocyte DNA. Using PBMC as a source, the number of T lymphocytes can be estimated at approximately 30% of total cells. Alternatively, the total nucleic acid can be isolated from cells, including both genomic DNA and mRNA. In other embodiments, the cDNA is transcribed from the mRNA and is then used as a template for amplification.

Multiplexed quantitative PCR

Multiplexed quantitative PCR was performed as described herein and in Robins et al., 2009 Blood 114, 4099; Robins et al., 2010 Sci. Translat. Med. 2: 47ra64; Robins et al., 2011 J. Immunol. Meth. doi: 10.1016 / j.jim.2011.09. 001; Sherwood et al. 2011 Sci. Translat. Med. 3: 90ra61; US 2012/0058902, US 2010/0330571, WO / 2010/151416, WO / 2011/106738 (PCT / US2011 / 026373), US 2015/0299785, WO2012 / 027503 (PCT / US2011 / 049012), US 2013 / 0288237, US 9,181,590, US 9,181,591 and US 2013/0253842. Current methods involve a multiplexed PCR method using a set of forward primers that specifically hybridize to V segments and a set of reverse primers that specifically hybridize to J segments of a TCR or Ig locus, wherein a PCR reaction Multiplexed using the primers allows the amplification of all possible combinations of VJ (and VDJ) within a given population of T or B lymphocytes.

Examples of V segment and J segment primers are described in US2012 / 0058902, US 2010/033057, WO / 2010/151416, WO / 2011/106738 (PCT / US2011 / 026373), US 2015/0299785, WO2012 / 027503 (PCT / US2011 / 049012), US 2013/0288237, US 9,181,590, US 9,181,591, US 2013/0253842, WO 2013/188831 (PCT / US2013 / 045994).

A multiplexed PCR system can be used to amplify the rearranged adaptive immune cell receptor loci. In certain embodiments, the CDR3 region is amplified from a CDR3 region of TCRA, TCRB, TCRG, or TCRD or similarly from an IgH or IgL locus (lambda or kappa). A plurality of V segment and J segment primers are used to amplify substantially all (e.g., greater than 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, or 99%) the adaptive immune receptor CDR3 encoding regions rearranged to produce a multiplicity of amplified rearranged DNA molecules. In certain embodiments, the primers are designed so that each amplified rearranged DNA molecule is less than 600 nucleotides in length, thus excluding amplification products from non-rearranged adaptive immune receptor loci.

In some embodiments, two sets of primers are used in a single highly multiplexed PCR reaction. The set of "forward" primers can include a plurality of oligonucleotide primers from segment V and the reverse set can include a plurality of oligonucleotide primers from segment J. In some embodiments, there is a primer that is specific (eg, having a nucleotide sequence complementary to a unique sequence region of) each segment of the V region and for each segment of the J region at the respective TCR or Ig gene locus. In other embodiments, a primer can hybridize to one or more V segments or J segments, thus reducing the number of primers required in multiplexed PCR. In certain embodiments, the J segment primers are combined with a conserved sequence in the junction segment ("J").

Each primer can be designed so that a respective amplified DNA segment is obtained that includes a portion of sequence of sufficient length to identify each J segment unambiguously based on sequence differences between known J region-encoding gene segments in the human genome database, and also to include a portion of sequence to which a J-segment specific primer can hybridize for resequencing. This design of V and J segment specific primers allows direct observation of a large fraction of somatic rearrangements present in the adaptive immune receptor gene repertoire within an individual.

In one embodiment, the present disclosure provides a plurality of V segment primers and a plurality of J segment primers. The plurality of V segment primers and the plurality of J segment primers amplify all or substantially all combinations of V segments. and J from a rearranged immune receptor locus. In some embodiments, the method provides amplification of substantially all rearranged AIR sequences in a lymphoid cell and is capable of quantifying the diversity of the TCR or IG repertoire of at least 106, 105, 104, or 103 unique rearranged AIR sequences in a sample. . "Substantially all combinations" can refer to at least 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99% or more of all combinations of V and J segments of a rearranged immune receptor locus. In certain embodiments, the plurality of V segment primers and the plurality of J segment primers amplify all combinations of the V and J segments of a rearranged immune receptor locus.

In general, a multiplexed PCR system can use 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25, and in certain embodiments, at least 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, or 39, and in other embodiments 40, 41 , 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 65, 70, 75, 80, 85 or more forward primers, wherein each forward primer specifically hybridizes or is complementary to a sequence corresponding to one or more segments of the V region. The multiplexed PCR system also uses at least 2, 3, 4, 5, 6, or 7, and in certain embodiments, 8, 9, 10, 11, 12, or 13 reverse primers, or 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25 or more primers, wherein each reverse primer specifically hybridizes to or is complementary to a sequence corresponding to one or more J region segments. Various combinations of V and J segment primers can be used to amplify common diversity. Complete set of TCR and IG sequences in a repertoire. For details on the multiplexed PCR system, including primer oligonucleotide sequences to amplify the TCR and IG sequences, see, eg, Robins et al., 2009 Blood 114, 4099; Robins et al., 2010 Sci. Translat. Med. 2: 47ra64; Robins et al., 2011 J. Immunol. Meth. doi: 10.1016 / j.jim.2011.09. 001; Sherwood et al. 2011 Sci. Translat. Med. 3: 90ra61; US 2012/0058902, US 2010/033057, WO / 2010/151416, WO / 2011/106738 (PCT / US2011 / 026373), US 2015/0299785, WO2012 / 027503 (PCT / US2011 / 049012), US 2013/0288237, US 9,181,590, US 9,181,591, US 2013/0253842, WO 2013/188831 (PCT / US2013 / 045994).

Oligonucleotides or polynucleotides that are capable of specifically hybridizing or hybridizing to a target nucleic acid sequence by nucleotide base complementarity can do so under moderate to high stringency conditions. In one embodiment, moderate to high stringency conditions suitable for specific PCR amplification of a target nucleic acid sequence may be between 25 and 80 PCR cycles, with each cycle consisting of a denaturation step (e.g., approximately 10-30 second (s) at greater than about 95 ° C), a hybridization step (for example, from about 10-30 s to about 60-68 ° C), and an extension step (for example, from about 10 -60 s to about 60-72 ° C), optionally in accordance with certain embodiments with the hybridization and extension steps combined to provide a two-step PCR. As would be recognized by the person skilled in the art, other PCR reagents can be added or changed in the PCR reaction to increase the specificity of the hybridization and amplification of the primer, such as altering the magnesium concentration, optionally adding DMSO, and / or the use of blocked primers, modified nucleotides, peptide nucleic acids, and the like.

A primer is preferably single-stranded DNA. The appropriate length of a primer depends on the intended use of the primer, but typically ranges from 6 to 50 nucleotides, or in certain embodiments, 15-35 nucleotides in length. Short primer molecules generally require cooler temperatures to form sufficiently stable hybrid complexes with the template. A primer need not reflect the exact nucleic acid sequence of the template, but must be complementary enough to hybridize to the template. The design of suitable primers for the amplification of a given target sequence is well known in the art and is described in the literature cited herein.

In some embodiments, the V and J segment primers are used to produce a plurality of amplicons from the multiplexed PCR reaction. In certain embodiments, the amplicons range in size from 10, 20, 30, 40, 50, 75, 100, 200, 300, 400, 500, 600, 700, 800, or more nucleotides in length. In preferred embodiments, the amplicons are between 50-600 nucleotides in length.

According to non-limiting theory, these embodiments take advantage of the current understanding in the art (also described above) that once an adaptive immune cell (e.g., a T or B lymphocyte) has rearranged its genes encoding the adaptive immune receptor (e.g. For example, TCR or Ig genes), their progeny cells possess the same rearrangement of the gene encoding the adaptive immune receptor, thus giving rise to a clonal population (clones) that can be uniquely identified by the presence in it of gene segments Rearranged V and J (eg, coding for CDR3) that can be amplified by a specific combination of V and J specific oligonucleotide primer pairs as described herein.

In some embodiments, segment V primers and segment J primers each include a second sequence at the 5 'end of the primer that is not complementary to the target V or J segment. The second sequence may comprise an oligonucleotide having a sequence that is selected from (i) a universal adapter oligonucleotide sequence, and (ii) a sequencing platform specific oligonucleotide sequence that is attached to and 5 'to a first sequence. oligonucleotide of the universal adapter. Examples of universal linker oligonucleotide sequences can be the forward pGEX and reverse pGEX linker sequences.

In some embodiments, the resulting amplicons using the V segment and J segment primers described above include amplified V and J segments and the universal adapter oligonucleotide sequences. The universal adapter sequence can be complementary to an oligonucleotide sequence found on a tail primer. The tail primers can be used in a second PCR reaction to generate a second set of amplicons. In some embodiments, the tail primers can have the general formula:

5-P-S-B-U-3 '(III),

wherein P comprises a sequencing platform specific oligonucleotide,

wherein S comprises an oligonucleotide sequence containing a sequencing platform tag;

wherein B comprises an oligonucleotide barcode sequence and wherein said oligonucleotide barcode sequence can be used to identify a sample source, and

wherein U comprises a sequence that is complementary to the oligonucleotide sequence of the universal adapter or is the same as the oligonucleotide sequence of the universal adapter.

Additional description on universal adapter oligonucleotide sequences, barcodes and tail primers are found in WO 2013/188831 (PCT / US13 / 45994).

Amplification bias control

Multiplexed PCR assays can lead to a bias in the total number of amplicons produced from a sample, as certain primer sets are more efficient at amplifying than others. To overcome the problem of such biased use of amplification primer subpopulations, methods that provide a template composition can be used to standardize the amplification efficiencies of the members of an oligonucleotide primer set, where the primer set is capable of amplifying Rearranged DNA encoding a plurality of adaptive immune receptors (TCR or Ig) in a biological sample comprising DNA from lymphoid cells.

In some embodiments, a template composition is used to standardize the various amplification efficiencies of the primer sets. The template composition may comprise a plurality of various template oligonucleotides of general formula (I):

5'-U1-B1-V-B2-R-J-B3-U2-3 '(I)

The constituent template oligonucleotides are diverse with respect to the nucleotide sequences of the individual template oligonucleotides. Individual template oligonucleotides can vary considerably in nucleotide sequence from one another based on significant sequence variability between the large number of possible polynucleotides of the variable region (V) and the TCR or BCR binding region (J). The sequences of individual template oligonucleotide species may also vary from each other based on sequence differences in U1, U2, B (B1, B2 and B3) and R oligonucleotides that are included in a particular template within the diverse plurality of molds.

In certain embodiments, V is a polynucleotide comprising at least 20, 30, 60, 90, 120, 150, 180, or 210, and no more than 1000, 900, 800, 700, 600, or 500 contiguous nucleotides of the sequence of a gene encoding an adaptive immune receptor variable region (V), or the complement thereof, and in each of the plurality of oligonucleotide sequences of the V template comprises a unique oligonucleotide sequence.

In some embodiments, J is a polynucleotide comprising at least 15-30, 31-60, 61-90, 91-120, or 120-150, and no more than 600, 500, 400, 300, or 200 contiguous nucleotides of the sequence of a gene encoding a junction region (J) or the complement thereof, and in each of the plurality of oligonucleotide sequences of template J comprises a unique oligonucleotide sequence.

U1 and U2 may each be none or each comprises an oligonucleotide independently having a sequence that is selected from (i) a universal adapter oligonucleotide sequence, and (ii) a sequencing platform specific oligonucleotide sequence that it is attached and positioned 5 'to the universal adapter oligonucleotide sequence.

B1, B2 and B3 can be none or each comprises an oligonucleotide B comprising a first and a second barcode sequence of oligonucleotides of 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900 or 1000 contiguous nucleotides (including all integer values between them), wherein each of the plurality of template oligonucleotide sequences B comprises a unique oligonucleotide sequence in which (i) the first code sequence of bars uniquely identifies the unique oligonucleotide V sequence of the template oligonucleotide and (ii) the second barcode sequence uniquely identifies the unique oligonucleotide J sequence of the template oligonucleotide.

R can be none at all or comprises a restriction enzyme recognition site comprising an oligonucleotide sequence that is absent from V, J, U1, U2, B1, B2, and B3.

The methods are used with the template composition to determine the non-uniform nucleic acid amplification potential between members of a set of oligonucleotide amplification primers that are capable of amplifying productively rearranged DNA encoding one or a plurality of adaptive immune receptors. in a biological sample comprising DNA from lymphoid cells from a subject. The method may include the steps of: (a) amplifying DNA from a template composition to standardize the efficiency of amplification of an oligonucleotide primer set in a multiplexed polymerase chain reaction (PCR) comprising: (i) the composition of the template (I) described above, wherein each template oligonucleotide in the plurality of template oligonucleotides is present in a substantially equimolar amount; (ii) a set of oligonucleotide amplification primers that is capable of productively amplifying rearranged DNA encoding one or a plurality of adaptive immune receptors in a biological sample comprising DNA from lymphoid cells from a subject.

The primer set may include: (1) in substantially equimolar amounts, a plurality of oligonucleotide V segment primers that are independently capable of specifically hybridizing to at least one polynucleotide encoding an adaptive immune receptor V region polypeptide or to the complement thereof, wherein each V segment primer comprises a nucleotide sequence of at least 15 contiguous nucleotides that is complementary to at least one functional gene segment encoding the functional adaptive immune receptor V region and wherein the plurality of primers segment V specifically hybridizes to virtually all adaptive immune receptor V region-encoding gene segments that are present in the template composition, and (2) in substantially equimolar amounts, a plurality of J segment oligonucleotide primers that are independently capable to hybridize specific with at least one polynucleotide encoding an adaptive immune receptor J region polypeptide or complement thereof, wherein each J segment primer comprises a nucleotide sequence of at least 15 contiguous nucleotides that is complementary to at least one functional gene segment that encodes the J region of the functional adaptive immune receptor and wherein the plurality of primers of the J segment specifically hybridizes with practically all the gene segments that encode the J region of the adaptive immune receptor that are present in the template composition.

Segment V and segment J oligonucleotide primers are capable of promoting amplification in said multiplexed polymerase chain reaction (PCR) of substantially all template oligonucleotides in the template composition to produce a multiplicity of amplified template DNA molecules. , said multiplicity of amplified template DNA molecules being sufficient to quantify the diversity of template oligonucleotides in the template composition, and wherein each amplified template DNA molecule in the multiplicity of amplified template DNA molecules is less than 1000, 900, 800, 700, 600, 500, 400, 300, 200, 100, 90, 80, or 70 nucleotides in length.

The method also includes the steps of: (b) sequencing all or a sufficient portion of each of said multiplicity of amplified template DNA molecules to determine, for each template DNA molecule unique in said multiplicity of template DNA molecules amplified, (i) a template-specific oligonucleotide DNA sequence and (ii) a relative frequency of occurrence of the template oligonucleotide; and (c) comparing the relative frequency of occurrence for each unique template DNA sequence of said template composition, wherein a non-uniform frequency of occurrence for one or more template DNA sequences indicates potential for non-uniform nucleic acid amplification. between the members of the oligonucleotide amplification primer set.

Additional description on bias control methods is provided in US 2013/0253842, US 9,150,905, US 2015/0203897 and WO 2013/169957 (PCT / US2013 / 040221).

Sequencing

Sequencing can be performed using any of a variety of available high throughput single molecule sequencing machines and systems. Illustrative sequence systems include sequence-by-synthesis systems, such as the Illumina Genome Analyzer and associated kits (Illumina HiSeq) (Illumina, Inc., San Diego, CA), the Helicos Genetic Analysis System (Helicos BioSciences Corp. , Cambridge, MA), Pacific Biosciences PacBio RS (Pacific Biosciences, Menlo Park, CA) or other systems with similar capabilities. Sequencing is accomplished using a set of sequencing platform specific oligonucleotides that hybridize to a defined region within the amplified DNA molecules. The sequencing platform specific oligonucleotides are designed to sequence amplicons, so that the gene segments encoding V and J can be uniquely identified by the sequences that are generated. See, for example, US 2012/0058902; US 2010/033057; WO 2011/106738 (PCT / US2011 / 026373); US 2015/0299785; or WO 2012/027503 (PCT / US2011 / 049012).

In some embodiments, the raw stream data is pre-processed to debug the primary stream of each read and compress the data. A nearest neighbor algorithm can be used to collapse data into unique sequences by fusing closely related sequences to eliminate sequencing and PCR errors. See, for example, US 2012/0058902; US 2010/033057; WO 2011/106738 (PCT / US2011 / 026373); US 2015/0299785; or WO 2012/027503 (PCT / US2011 / 049012).

Estimation of the abundance of the PCR template

To estimate the average read coverage per input template in the multiplexed PCR and sequencing approach, a set of synthetic TCR (or BCR) templates (as described above) can be used, comprising each combination of Vp gene segments. and Jp. These synthetic molecules can be those described in the general formula (I) above and in US 2013/0253842, US 9,150,905, US 2015/0203897 and WO 2013/169957 (PCT / US2013 / 040221).

These synthetic molecules can be included in each PCR reaction at a very low concentration so that only some of the synthetic templates are seen. Using the known concentration of the set of synthetic templates, the relationship between the number of unique synthetic molecules observed and the total number of synthetic molecules added to the reaction can be simulated (this is almost one to one at the low concentrations that were used). Synthetic molecules allow the average number of sequencing reads obtained per PCR template molecule and an estimate of the number of T lymphocytes in the input material carrying each unique TCR rearrangement to be calculated for each PCR reaction.

Discovery of diagnostic public T cell responses

Given a large population of subjects with and without a particular infectious disease, the responses of diagnostic public T cells to disease state can be determined by applying the following statistical method. An immune receptor repertoire of unique T-cell receptor sequences is determined, using the methods described above, for a group of subjects who have been identified as having or not having an infectious disease, such as CMV or smallpox. It is then possible to determine which of these T cell receptor sequences are significantly more common (ie, present in more individuals) among subjects with the disease state than in subjects without the disease state. When a common TCR binds to an antigen, it is called the public T cell response. Public T cell responses are specific for a particular disease or antigen, are present in many individuals, and are encoded in a common format (specific rearranged receptor sequences) regardless of disease.

A one-tailed Fisher's exact test is used, using the presence or absence of the T cell receptor sequence in question versus with or without disease status to construct a 2x2 contingency table (shown below) and with PDT (false discovery rate) controlled using an empirical distribution of null p-values determined by disease state permutations. Since many clones are unique to a single subject (and consequently unique to positive or negative disease state classes), it is vital to control the rate of false discoveries in trait selection to avoid over-fitting to the many False associations of unique TCRp with positive infectious disease status. This process generates a list of T cell receptor sequences that are significantly more common in subjects with the disease state of interest.

Use of public T-cell diagnostic responses to infer disease status

Given a large population of subjects with and without a given infectious disease, and given a list of diagnostic public T-cell responses generated as described above, it is possible to infer disease status in a subject whose disease status is unknown. .

First, a "burden of disease" or quantitative measure of the presence and / or abundance of such diagnostic public T cell responses is calculated for each subject (those with known status as well as the subject with unknown status). Burden of disease or quantitative measure is the proportion of unique T-cell receptor sequences in each subject that are among the list of diagnostic public T-cell responses.

Once this measure is calculated, it is then determined whether the subject of unknown status has a disease burden consistent with subjects known to have the disease status of interest or with subjects known not to have the disease status of interest. disease of interest. Here, the comparison method is to train a logistic regression model of (burden of disease versus presence of disease state) in all subjects with known disease state, and use that model to assign the subject to unknown state a probability of having the disease state of interest.

In some embodiments, the method comprises a classification model that uses extensive cross-validation of leaving one out, in which an individual's immune profile data set is excluded from the calculations and the process is repeated from the beginning. In one aspect, cross-validation is used to evaluate the precision of a classification model. The resulting classification model is used to predict the infectious disease status of the subject who has been left out.

These methods can be performed to determine disease status for various types of disease, including, but not limited to, CMV, EBV, HPV, HIV, smallpox, other infectious diseases, or even non-infectious diseases, such as autoimmune diseases and neoplasms.

HLA typing

HLA typing of bone marrow donor subjects was performed according to the standard protocol of the Fred Hutchinson Cancer Research Center, and can be performed by methods known to those of skill in the art.

Method for determining the HLA status of unknown subjects

FIG. 13 shows an overview of the method for inferring HLA status. This method can be applied to any type of HLA allele and to any sequence profile of TCRs, including the gene sequences of TCRA, TCRB, TCRG or TCRD.

First, the type of HLA allele is determined for each subject in a group, according to methods standard in the art for HLA typing. Next, a locus from the subject's immune repertoire is sequenced. For example, the TCRB locus is amplified and sequenced, using the methods described above.

For each subject, the total number of unique TCRB sequences and the frequency of each unique TCRB sequence are determined. For a given HLA allele, it is determined which TCRB sequences are significantly associated with the HLA allele. In one embodiment, for each unique TCRB sequence, it is determined how many subjects who are positive for an HLA allele have the TCRB sequence and how many subjects who are negative for the HLA allele have the TCRB sequence. In addition, the number of subjects that are positive for an HLA allele and negative for the TCRB sequence and the number of subjects that are negative for the h La allele and positive for the TCRB sequence can be determined. The following table shows the categorization of subjects according to the presence or absence of a TCRB sequence and the presence or absence of an HLA allele, HLA-A2.

FIG. 14 shows a listing of illustrative unique TCRB sequences and the number of subjects who are positive or negative for an HLA-A2 allele who have a particular TCRB sequence.

A p-value is determined for the association of each TCRB sequence with an HLA status using a Fisher's exact (two-tailed) test. The p-value for the association of each TCR with allele status using Fisher's exact test (two tailed) is calculated as follows:

FIG. 14 also shows illustrative p-values that are calculated for the association of a particular TCRB sequence with an HLA type (HLA-A2).

As shown in FIG. 15, a p-value is selected as a cut-off point to identify a set of "TCR traits" from the complete list of possible TCR sequences. Defining a threshold of p-value and permutation of allele status between individuals provides an estimate of the false discovery rate. This is done for each HLA allele, resulting in a set of allele-associated TCRp sequences for each HLA allele present in the training data. In FIG. 15, a cutoff value of p-value of p <10-4 and a TDF of 0.1 are used to identify 288 TCRB sequences that are positively associated with HLA-A2. For each of the TCRB sequences associated with the alleles, the frequency of the sequence is also determined in each subject.

The trait selection stage is followed by a machine learning process. As shown in FIG. 16, for each HLA allele, a logistic regression model is trained using the set of characteristic vectors on all subjects, along with the known status of the presence of that allele.

As shown in FIG. 17, a comprehensive leave-one-out cross-validation is performed where a subject is removed from analysis, and the subject's HLA status is inferred based on trait selection and training of only the remaining subjects. The result is a set of classifiers (one for each HLA allele) that estimate the probability of a positive status for each HLA allele, taking as input the vector of characteristics for each allele.

Equipped with these classifiers, the HLA type of a new subject can be assessed by: 1. sequencing the immune repertoire, 2. calculating the trait vectors for each allele, and 3. defining a probability threshold for the positive state to be will apply to the output of each classifier. In this way, an HLA status can be inferred for a new subject with an unknown HLA status.

In one embodiment, given the results of the trait selection scheme, a one-dimensional allelic score is determined as the fraction of an individual's unique TCRp sequences that appear in the set of TCRp sequences associated with a given allele. This single quantity, rather than a vector of traits, can also be used to train a classifier and diagnose new individuals with similar precision. Additionally, the specific classifier described (logistic regression) can be replaced with any of the other binary classifiers and gives substantially similar results. These include: k-nearest neighbors, random forests, neural network! artificial, naive Bayes and support vector machine.

Examples

The following examples are offered for illustrative purposes only and are not intended to limit the scope of the present invention in any way. Efforts have been made to ensure accuracy with respect to the numbers used (eg quantities, temperatures, etc.), but of course some experimental errors and deviations should be allowed.

The practice of the present invention will employ, unless otherwise indicated, conventional methods of protein chemistry, biochemistry, recombinant DNA techniques, and pharmacology, within the skill of the art. These techniques are fully explained in the literature. See, for example, TE Creighton, Proteins: Structures and Molecular Properties (WH Freeman and Company, 1993); AL Lehninger, Biochemistry (Worth Publishers, Inc., current addition); Sambrook, et al., Molecular Cloning: A Laboratory Manual (2nd Edition, 1989); Methods In Enzymology (S. Colowick and N. Kaplan eds., Academic Press, Inc.); Remington's Pharmaceutical Sciences, 18th Edition (Easton, Pennsylvania: Mack Publishing Company, 1990); Carey and Sundberg Advanced Organic Chemistry 3rd Ed. (Plenum Press) Vol. A and B (1992).

Example 1: Identification of TCRs Associated with Cytomegalovirus (CMV) and Classification of CMV Status from Immunosequencing Data

The first goal of the study was to create a comprehensive catalog of CMV-specific T-cell receptor sequences. The objective was to identify common and rare shared TCRp sequences enriched in CMV-seropositive individuals relative to CMV-seronegative individuals. CMV was used as model because it provides an ideal test bed for the development of T cell biomarkers. CMV serostatus is widely available and easy access to nearly equal numbers of seropositive (cases) and seronegative (control) individuals among healthy adults provides a good statistical power for rigorous analysis.

There are many possible TCR sequences generated during VDJ recombination (~ 1012 easily accessible via VDJ recombination). As shown in FIG. 1, different VDJ recombination events can result in the same nucleotide sequence. Furthermore, it is possible that VDJ recombination may result in different nucleotide sequences that are translated into the same amino acid sequence. See Venturi, Price, Douek, and Davenport 2008. The molecular basis for public T-cell responses? Nature Reviews Immunology. Certain TCRs are common rearrangements shared by more than one individual. When a common TCR binds to an antigen, it is called the public T cell response. Public T cell responses are specific for a particular disease or antigen, are present in many individuals, and are encoded in a common format (specific rearranged receptor sequences) regardless of disease.

In this study, the response of public T cells to cytomegalovirus (CMV) was examined by sequencing T-cell rearranged receptor (TCR) genes obtained from 640 subjects. CMV is a lifelong, usually asymptomatic infection that affects most adults and elicits a robust memory T-cell response.

FIGS. 2A, 2B and 2C provide a general description of the method, in accordance with one embodiment of the invention. FIG. 2A shows a data set of peripheral blood samples from 640 healthy subjects (287 CMV- and 353 CMV +), which were analyzed using high-throughput TCR profiles. In FIG. 2B, unique TCRp sequences were identified that were present in significantly more CMV + subjects than CMV- subjects, controlling for the false discovery rate (TDF) per permutation of CMV status. The presence of these CMV-associated TCRp sequences was used to construct a classification model. In FIG. 2C, the classification model was tested using extensive cross-validation of leaving one out, in which a sample was kept out of the calculations and the process was repeated from the beginning. The resulting classification model was used to predict the CMV status of the excluded subject.

Human peripheral blood samples were obtained from the healthy bone marrow donor biorepository of the Fred Hutchinson Cancer Research Center Research Cell Bank under a protocol following written informed consent approved and supervised by the Fred Hutchinson Cancer Research Center Institutional Review Board. This biorepository houses an inventory of PBMC, B-LCL, and DNA from hematopoietic cell transplant (HCT) patients, donors, and family members along with cell lines and DNA sourced from selected stem cell transplant patients, donors, and family members. of ~ 7,800 stem cell transplant patients and 7,300 donors (from whom samples were drawn). DNA for this study was drawn from peripheral blood of HCT donors and extensively typed for HLA antigens and alleles along with other defining phenotypes required for donor-patient matching or donor inclusion / exclusion. These include cytomegalovirus (CMV), Epstein-Barr virus (EBV), herpes simplex virus (HSV), hepatitis, and diabetes mellitus.

As shown in FIG. 3A, 640 subjects were phenotyped specifically for HLA type and CMV status to be eligible as TCH donors, and they were separated into 287 ^c M ^v donors and 353 CMV donors. The seropositive rate for CMV was approximately 45% (approximately the same number of seropositive and seronegative samples for the investigation of public T cell responses).

FIG. 3B shows the demographic characteristics of the subjects included in this study, classified by CMV status.

S e c u e n c i a t i o n d e T C R B de al to r e n d i m i e n t:

Genomic DNA was extracted from peripheral blood samples using the Qiagen DNeasy Blood Collection Kit (Qiagen, Gaithersburg, MD, USA). The CDR3 regions of the rearranged TCR-p genes were sequenced; the CDR3 region of TCR-p was defined according to the collaboration IMGT28,20. The CDR3 regions of TCRp were amplified and sequenced using the methods described above and in the protocols described above5,29. The multiplexed PCR method used a mixture of 60 specific forward primers for the TCR Vp gene segments and 13 specific reverse primers for the TCR Jp gene segments. The resulting amplicons were sequenced using the methods described above. 87 bp readings were obtained using the Illumina HiSeq system. The raw HiSeq sequence data was pre-processed to eliminate errors in the main sequence of each read and to compress the data. A nearest neighbor algorithm was used to collapse the data into unique sequences by fusing closely related sequences, to eliminate sequencing and PCR errors.

In order to ensure adequate coverage of each T lymphocyte rearrangement, a coverage of 8-10x sequence, or ~ 6-10 million sequencing reads per sample, using approximately eight full sequencing runs. All sequencing reads were processed using a standardized bioinformatics model under development to 1) demultiplex reads for specific samples, 2) remove low quality sequences and remove potential contaminants, 3) align and identify specific gene segments from CDR3 regions of V and J from TCRp, 4) cluster highly similar sequences to account for PCR and sequencing errors, 5) normalize data to eliminate PCR amplification bias, 6) estimate total T lymphocyte input, and 7) generate unique sequence counts and distributions of TCRp.

Herein, approximately 250,000 rearranged T-cell receptor genes were sequenced from the peripheral blood of each subject.

S e l e c t i o n d e r a s a o s:

In 640 subjects, there were 185,204 (+/- 84,171) unique TCRp per subject and 83,727,796 unique TCRp in total. Rather than attempting a high-dimensional CMV classification using all single TCRp as possible traits, a new trait selection scheme was developed. Trait selection was the selection of a particular amino acid sequence to identify common TCRs among individuals with CMV.

Since many clones were unique to a single subject (and consequently unique to the CMV + or CMV- classes), it was vital to control the rate of false discoveries in the selection of traits to avoid an over-adjustment to the many spurious associations of Unique TCRp with CMV status.

Each unique TCRp rearrangement, identified by V and J gene assignment and CDR3 amino acid sequence, was tested for CMV association. Each of these underwent a one-tailed Fisher's exact test to determine their incidence in CMV- and CMV + subjects. Specifically, letting nj denote the number of subjects with CMV status j (with j - o) and clone i present, a p-value was calculated, performing Fisher's exact test on the contingency table.

where N + and N. denote the total number of subjects with each CMV status (CMV + and CMV-, respectively).

To characterize a rejection region in the presence of many weakly dependent hypotheses (one for each unique TCRp), CMV status assignments were randomly permuted 100 times and statistics on the number of rejections at the theoretical threshold of p-value were calculated. . Approximating the total fraction of true null hypotheses as a unit, this allowed estimation of the false discovery rate (TDF) as the ratio between the mean number of rejections under permutation and the actual number of rejections.

As shown for example in FIG. 4, the amino acid sequence of TCR "CASSLIGVSSYNEQFF" (SEQ ID NO: 12) was selected and identified in 27 CMV + subjects and 2 CMV- subjects (p = 2.8E-08).

FIGS. 5 shows illustrative TCR amino acid sequences identified in CMV + subjects. As shown in Table 1, at a selected p-value of 10-4, 142 public T cell clones were identified as associated with CMV + subjects. The p-value was chosen at 10-4, but as seen, at a lower p-value, a greater number of CMV-associated TCRB sequences can be identified.

Table 1: Each of the 142 TCRp sequences significantly associated with CMV status (p <1x10-4, TDF ~ 20%) was included in the cohort. The sequences were defined using the amino acid sequence of the CDR3 region together with the V and J gene segments. The number of CMV + and CMV- subjects in which each sequence was observed is given, as well as the p-value of an exact test. Fisher's for association with CMV status, and any HLA-A or HLA-B allele that was significantly associated (p <1x10-3) with the presence of each sequence.

_ _ _ _ R eduction of dimensionality and automatic rendering:

The CMV load was calculated for each subject. CMV loading is defined as the fraction of a subject's unique TCRps that are significantly associated with CMV, as shown in FIG. 6. The graph of FIG. 6 shows the distribution of CMV scores (that is, the proportion of each subject's TCRp repertoire that matches the listing of 142 CMV-associated TCRp sequences) between CMV + and CMV- subjects.

This allowed for rapid training of a one-dimensional logistic regression classifier of CMV status. Extensive one-out cross-validation (including recalculation of CMV-associated clones) was performed, indicating high precision across a wide range of p-value thresholds, with performance degraded at high TDF.

Each subject was removed from the data set in turn, and the list of TCRp significantly associated with CMV was recalculated using the remaining subjects with known CMV status. The CMV load of each subject was then calculated as described above. Next, a one-dimensional logistic regression classifier was trained on CMV load versus CMV status, and the subject's CMV load kept out was calculated and entered into the logistic regression. The result of this logistic regression model was the probability that the subject of unknown status was CMV +, which was then compared to the status (known but not potentiated) of the excluded subject to determine the precision of the exclusion classification. The classification of a subject with a truly unknown CMV status proceeds in the same way: sequencing the TCRp, calculating a CMV load (as a proportion of unique TCRp of this subject present in the list of TCRp associated with CMV) and entering this load of CMV in a trained logistic regression model in subjects of known status, with an estimated probability of CMV positivity as the output.

As shown in FIG. 7, the cross-validation method left out one of the initial 640 samples, then the CMV-specific TCRp sequence database and associated statistics were retrained in an unsupervised manner to eliminate bias, and classified the CMV serological status of the excluded sample.

The cross-validation results are shown in FIGS. 8 and 9 (displayed for all subjects and cross-validated (CV) subjects). The accuracy of the model was shown to be 89%, corresponding to a diagnostic chance ratio of ~ 66. By evaluating the performance of the model in the range of specificity versus sensitivity using a COR curve, an area under the curve was achieved. 93% COR.

FIG. 8 (top graph) shows data for the classification performance of all and cross-validation (CV) data sets for each threshold p-value, measured as the area under the COR curve (AUCOR). The number on each set of data points corresponds to the number of TCRp associated with CMV identified at that p-value threshold, and the rectangle indicates the data set selected for subsequent analysis (p-value = 10 "4) .

FIG. 8 (bottom graph) also shows an estimated false discovery rate (TDF) for each threshold p-value used in the identification of TCRp sequences significantly associated with CMV, using permutations of CMV status. The best performance is observed with a p-value of 10-4, which corresponds to an estimated TDF of ~ 20%, resulting in the identification of a set of 142 TCRp sequences that were significantly associated with CMV positive status. (listed in Table 1). The use of these conditions resulted in a good separation between CMV + and CMV- subjects in the cohort as measured by the CMV score (see FIG. 6).

FIG. 9 shows the COR curves for the total and cross-validation data sets. The AUCOR for the entire data set is 0.98, indicating that our approach resulted in an excellent classifier for CMV status. At the point of greatest discrimination power, a precision of 0.89 and a diagnostic probability ratio of 66 were observed in the cross-validation data set (achieved by correctly classifying 86% of true positives with a rate of 8% false positives). Taken together, these data suggest that the presence of public T cell responses to CMV is highly correlated with CMV positive status.

Since T lymphocytes recognize their cognate antigens in the environment of MHC molecules expressed by antigen presenting cells, it was tested whether HLA restriction of CMV-associated TCRp sequences of the present inventors could be identified. A Fisher's exact test was performed on each CMV-associated TCRp sequence to determine whether their presence was significantly associated with any of the HLA alleles observed in the cohort. The association of 57 of the 142 CMV-associated TCRp sequences could be confidently assigned to at least one HLA allele, with a p-value cutoff of 1x10-3. Full results are presented in Table 1 (above) and FIG. 10.

FIGS. 10A and 10B show HLA restriction of CMV-associated TCRp sequences. FIG. 10A shows the distribution of HLA-A alleles in this cohort. FIG. 10B shows the distribution of HLA-B alleles in this cohort. Each of the 142 CMV-associated TCRp sequences identified at p <1x10-4 was tested for a significant association with each HLA allele, with a threshold p-value of 1x10-3 Of these 142 associated TCRp sequences to CMV, 57 were significantly associated with one HLA-A and / or HLA-B allele, and no sequence was significantly associated with more than one allele at each locus. The colored boxes and amino acid sequences indicate the 5 sequences of the TCRp identified in the present inventors' study that had been previously identified. In 4 of the cases, the correct HLA association was recapitulated, and in the 5th case, there was no statistically significant HLA association.

A search was made in the literature to identify sequences of TCRP CMV reagents previously ^{TCRP I qe had been identified os} _E ^or

_st ^c

_or ^or

_d ^m

_io ^on

_p ^t

_r ^to

_and ^d

_v ^to

_ia ^s

_m ^{, and} J ^{I know}

_entity ^s

_p ^and

_or ^you

_b ^c

_li ^c

_c ⁱ

_to ^or

_d ⁿ

_or 3 ^ro

_' 1 ⁿ 0 ⁵ 1 ⁹ 2 ⁵ 27 ^{sequences d}

_{. Of these, 30 secu} 1 ^unique

_{gums only} 1 ^u

_{icas of the TCRp had s} 1 _i ^p

_d ^or

_{or re} ^ra _p ^{l e m} _e ⁿ _{v mind} classified as public or were documented in numerous studies. 5 of these 30 unique public TCRp sequences were found to be contained within the 142 CMV-associated TCRp set. Furthermore, an HLA association was identified for 4 of these 5 sequences of the TCRp, and in all 4 cases, the HLA association of the present inventors was in agreement with the literature. FIGS. 11A, 11B and 12 provide more complete information on the prevalence in our data set of previously identified CMV-reactive TCRp sequences, as described below.

FIGS. 11A and 11B show the incidence of previously documented CMV-reactive TCRp sequences in this cohort. After a literature search, 565 'private' CMV-reactive TCRp sequences (documented in one individual from a single study) and 30 'public' CMV-reactive TCRp sequences (documented in multiple studies or in multiple individuals were identified within a single study). FIG. 11A shows the incidence of each sequence of PCRRs in the cohort of 640 subjects represented graphically along the horizontal axis by decreasing the total incidence, with the incidence in CMV + subjects above the horizontal and the incidence in CMV- subjects. below horizontal. Many previously documented sequences were seen in the data set, but most were seen in roughly the same number of CMV + and CMV- subjects, which could be explained by receptor sequences with exceptionally high frequency in the repertoire without prior exposure, or it could reflect cross-reactive receptors that bind to CMV antigens, but also to other common antigens. FIG. 11B shows an incidence histogram of these sequences of the TCRp in the cohort of 640 subjects plotted for each group of sequences. Most of the previously documented CMV-reactive TCRp sequences were found in our data set at appreciable levels, although only a handful were found disproportionately in CMV + subjects at p <1x10-4. The "public" TrRp sequences described in the literature were considerably more common in our cohort.

FIG. 12 shows the concordance of the TCRB sequences in the cohort compared to those in the literature. Of the 142 TCRp sequences significantly associated with CMV status in the cohort with a p-value of less than 1x10-4, five TCRp sequences (defined by the coincidence of the CDR3 amino acid sequence) have been previously documented in the bibliography as public clones. FIG. 12 provides the amino acid sequence of CDR3, the V and J genes, and the HLA association for each sequence and compares it to previous reports. As expected, the two sequences previously identified as public were seen in more subjects in our cohort than the three sequences previously documented only once. The concordance of the V gene, the J gene and the HLA allele with those described in the bibliography was very good. Five of the nine total comparisons had the same V gene and the other 4 comparisons had the same subfamily of the V gene. There were concordant J genes in 9 of 9 comparisons, and an identical HLA association in 8 of 9 comparisons, with no sequence. significantly associated with HLA in this study.

Thus, immunosequencing can be used to determine CMV-specific T cell sequences in a large cohort and predict CMV status in new subjects with high precision. In summary, it has been shown that information obtained from rearranged T-cell receptors can be used to infer disease status based on the presence of public T-cell responses; the only requirement is a large sample of pathogen positive and negative samples with which to identify these public T cell responses. Because high-throughput sequencing of T-cell receptors captures all T-cell responses equally, and they store immunological memory for all pathogens in a common format, reading T-cell memory looking for Known public responses may be a viable strategy to simultaneously diagnose a wide range of infectious agents using a single peripheral blood sample and a single unitized assay. These methods are necessarily applicable to acute infections, since the memory of T lymphocytes persists for years, which explains the fact that it is not known how public clones decay over time after an acute infection. Accordingly, the present method can be used to evaluate multiple infections simultaneously, such as HPV, EBV, CMV, and others. The method can also be used to predict or diagnose a non-infectious (eg autoimmune) disease.

The methods described herein can also be used to quantify disease burden (eg, CMV reactivation after transplantation).

Example 2: Identifying the type of HLA

The first aim of the study was to create a comprehensive catalog of T cell receptor sequences and associated HLA types.

H L A T w i n a:

640 subjects were phenotyped for the HLA type according to the standard protocol of the Fred Hutchinson Cancer Research Center.

S e c u e n c i a t i o n de l r e p e r t o r i o i n m u n i t a r i o:

The TCRB sequences for each subject were then determined by amplification and sequencing.

Genomic DNA was extracted from peripheral blood samples using the Qiagen DNeasy Blood Collection Kit (Qiagen, Gaithersburg, MD, USA). The CDR3 regions of the rearranged TCRp genes were sequenced; the CDR3 region of TCRp was defined in collaboration with IMGT28, 20 The CDR3 regions of TCRp were amplified and sequenced using the methods described above and in the protocols described above.5, 29 The multiplexed PCR method used a mixture of 60 primers specific direct for the TCR Vp gene segments and 13 specific reverse primers for the TCR Jp gene segments. The resulting amplicons were sequenced using the methods described above. 87 bp readings were obtained using the Illumina HiSeq system. The raw HiSeq sequence data was pre-processed to eliminate errors in the main sequence of each read and to compress the data. A nearest neighbor algorithm was used to collapse the data into unique sequences by fusing closely related sequences, to eliminate sequencing and PCR errors.

To ensure adequate coverage of each T cell rearrangement, 8-10x sequence coverage, or ~ 6-10 million sequencing reads per sample was generated using approximately eight full sequencing runs. All sequencing reads were processed using a standardized bioinformatics model under development to 1) demultiplex reads for specific samples, 2) remove low quality sequences and remove potential contaminants, 3) align and identify specific gene segments from CDR3 regions of V and J from TCRp, 4) cluster highly similar sequences to account for PCR and sequencing errors, 5) normalize data to eliminate PCR amplification bias, 6) estimate total T lymphocyte input, and 7) generate unique sequence counts and distributions of TCRp.

Herein, approximately 250,000 rearranged T-cell receptor genes were sequenced from the peripheral blood of each subject. For each subject, approximately 105 TCRB sequences were obtained and the abundances of each unique sequence were quantified.

S e l e c c i o n d e r a s a o s: P a r a d e f i r la a s o c i a c i o n e n t r e u n a l e l o H L A y s e c u e n c i a s d e T C R B

For each subject, the total number of unique TCRB sequences and the frequency of each unique TCRB sequence are determined. For a single TCRB sequence, it is determined how many subjects who are positive for an HLA allele have the TCRB sequence and how many subjects who are negative for the HLA allele have the TCRB sequence. Furthermore, the number of subjects who are positive for an HLA allele and negative for the TCRB sequence and the number of subjects who are negative for the HLA allele and positive for the TCRB sequence can be determined. The following table shows the categorization of subjects according to the presence or absence of a TCRB sequence and the presence or absence of an HLA allele, HLA-A2.

FIG. 14 shows a listing of illustrative unique TCRB sequences and the number of subjects who are positive or negative for an HLA-A2 allele who have a particular TCRB sequence.

A p-value is determined for the association of each TCRB sequence with an HLA status using a Fisher's exact (two-tailed) test. The p-value for the association of each TCR with allele status using Fisher's exact test (two tailed) is calculated as follows:

FIG. 14 shows illustrative p-values that were calculated for the association of a particular TCRB sequence with an HLA type (HLA-A2).

As shown in FIG. 15, a p-value is selected as a cut-off point to identify a set of "TCR traits" from the complete list of possible TCR sequences. Defining a threshold of p-value and permutation of allele status between individuals provides an estimate of the false discovery rate. This is done for each HLA allele, resulting in a set of TCRp sequences associated with the alleles for each HLA allele. In FIG. 15, a cutoff value of p-value of p <10-4 and a TDF of 0.1 are used to identify 288 TCRB sequences that are positively associated with HLA-A2. For each of the TCRB sequences associated with the alleles, the frequency of the sequence is also determined in each subject.

As shown in FIG. 15, a p-value is selected as a cut-off point to identify a set of "TCR traits" from the complete list of possible TCR sequences. Defining a threshold of p-value and permutation of allele status between individuals provides an estimate of the false discovery rate. This is done for each HLA allele, resulting in a set of TCRp sequences associated with the alleles for each HLA allele. In FIG. 15, a cutoff value of p-value of p <10-4 and a TDF of 0.1 are used to identify 288 TCRB sequences that are positively associated with HLA-A2. For each of the TCRB sequences associated with the alleles, the frequency of the sequence is also determined in each subject.

The false discovery rate (TDF) was determined by permutation of the allele status. See Storey et al. Statistical significance for genomewide studies. PNAS, 100 (6), pp. 9440-9445.

The trait selection stage is followed by a machine learning process. As shown in FIG. 16, for each HLA allele, a logistic regression model is trained using the set of characteristic vectors on all subjects, along with the known status of the presence of that allele.

As shown in FIG. 17, a comprehensive leave-one-out cross-validation is performed where a subject is removed from analysis, and the subject's HLA status is inferred based on trait selection and training of only the remaining subjects. The result is a set of classifiers (one for each HLA allele) that estimate the probability of a positive status for each HLA allele, taking as input the vector of characteristics for each allele.

Equipped with these classifiers, the HLA type of a new subject can be assessed by: 1. sequencing the immune repertoire, 2. calculating the trait vectors for each allele, and 3. defining a probability threshold for the positive state to be will apply to the output of each classifier. In this way, an HLA status can be inferred for a new subject with an unknown HLA status.

FIG. 18 shows the results of a cross-validation study. The cross-validation of leaving one out was done for 78 rounds. For the HLA-A2 allele, the method described above resulted in the identification of 288 TCR traits and an accurate prediction of 41 out of 43 subjects as HLA-A2 positive. As described above, the HLA type of the subjects was known prior to cross-validation. Only 2/43 subjects were false positives for HLA-A2. Thus, the method was 96% accurate in predicting the presence of HLA-A2 in a subject based on the subject's TCRB sequence profile.

FIG. 18 also shows the results of the cross-validation of the HLA-A24 allele, which was also 96% accurate. 10 of 13 subjects were accurately predicted to have the HLA-A24 allele, and 65 of 65 subjects were accurately predicted not to have the HLA-A24 allele based on identified TCRB traits.

Example 3: HLA Study

In another example, the response of public T cells to cytomegalovirus (CMV) was investigated by sequencing rearranged T cell receptors (TCRs) in 650 subjects (294 with and 356 without CMV). The concordance between ~ 90 million unique CRTs and CMV serostatus was evaluated, focusing on the identification of significant associations. In this study, 157 CRTs associated with CMV ap <10-3, TDF “0.15) were identified. By training a binary classifier on these traits, CMV serostatus in a cross-validation procedure of leaving one out was predicted to have a diagnostic odds ratio of 44. The classifier was also tested in a second independent cohort of 120 subjects with known CMV serostatus, generating a diagnostic odds ratio of 49.

Next, the HLA restriction of each CMV-associated TCR was investigated by evaluating the over-representation of particular HLA types among subjects carrying each CMV-associated TCR. Of 157 CMV-associated TCRs, 61 were HLA-associated at p <10-3 None of these were significantly associated with multiple HLA-A or HLA-B alleles.

There was substantial agreement between our data and previously published TCRs associated with CMV and HLA. Most of the previously documented public CMV-specific TCRs were seen in the data, although only 5/157 CMV-associated TCRs identified in this study have been previously documented. Of these, 4 had significant HLA restriction and all four confirmed previous findings.

Furthermore, the association of TCRs with each HLA-A allele present in the cohort was investigated, with significant results (p <10 "4) for many higher frequency alleles. Training of the binary classifier resulted in a highly accurate prediction. (-96%) for these alleles, indicating that the HLA type can be inferred from the immunosequencing data.

In summary, the present study demonstrated the validity of association studies using immunosequencing for detection and HLA association of public T-cell responses to infection, and showed that assessing the presence of associated T-cell responses it can serve as a powerful diagnostic classifier.

REFERENCES

1. Neller, M.A., Burrows, J.M., Rist, M.J., Miles, J.J. & Burrows, S.R. High frequency of herpesvirus-specific clonotypes in the human T cell repertoire can remain stable over decades with minimal turnover. J Virol 87, 697 700 (2013).

2. Li, H., Ye, C., Ji, G. & Han, J. Determinants of public T cell responses. Cell Res 22, 33-42 (2012).

3. Venturi, V., Price, D.A., Douek, D.C. & Davenport, M.P. The molecular basis for public T-cell responses? Nat Rev Immunol 8, 231-238 (2008).

4. Arstila, TP et al. A direct estimate of the human alphabeta T cell receptor diversity. Science 286, 958-961 (1999).

5. Robins, HS et al. Comprehensive assessment of T-cell receptor beta-chain diversity in alphabeta T cells. Blood 114, 4099-4107 (2009).

6. Cabaniols, J.P., Fazilleau, N., Casrouge, A., Kourilsky, P. & Kanellopoulos, J.M. Most alpha / beta T cell receptor diversity is due to terminal deoxynucleotidyl transferase. J Exp Med 194, 1385-1390 (2001).

7. Davis, M.M. & Bjorkman, P.J. T-cell antigen receptor genes and T-cell recognition. Nature 334 (1988).

8. Robins, HS et al. Overlap and effective size of the human CD8 + T cell receptor repertoire. Sci Transl Med 2, 47ra64 (2010).

9. Li, H. et al. Recombinatorial biases and convergent recombination determine interindividual TCRbeta sharing in murine thymocytes. J Immunol 189, 2404-2413 (2012).

10. Venturi, V. et al. TCR beta-chain sharing in human CD8 + T cell responses to cytomegalovirus and EBV. J Immunol 181, 7853-7862 (2008).

11. Gandhi, M.K. & Khanna, R. Human cytomegalovirus: clinical aspects, immune regulation, and emerging treatments. Lancet Infect Dis 4, 725-738 (2004).

12. Babel, N. et al. Clonotype analysis of cytomegalovirus-specific cytotoxic T lymphocytes. J Am Soc Nephrol 20, 344-352 (2009).

13. lancu, EM et al. Clonotype selection and composition of human CD8 T cells specific for persistent herpes viruses varies with differentiation but is stable over time. J Immunol 183, 319-331 (2009).

14. Khan, N., Cobbold, M., Keenan, R. & Moss, P.A. Comparative analysis of CD8 + T cell responses against human cytomegalovirus proteins pp65 and immediate early 1 shows similarities in precursor frequency, oligoclonality, and phenotype. J Infect Dis 185, 1025-1034 (2002).

15. Klarenbeek, PL et al. Deep sequencing of antiviral T-cell responses to HCMV and EBV in humans reveals a stable repertoire that is maintained for many years. PLoSPathog 8, e1002889 (2012).

16. Klinger, M. et al. Combining next-generation sequencing and immune assays: a novel method for identification of antigen-specific T cells. PLoS One 8, e74231 (2013).

17. Koning, D. et al. In vitro expansion of antigen-specific CD8 (+) T cells distorts the T-cell repertoire. J Immunol Methods 405, 199-203 (2014).

18. Miconnet, I. et al. Large TCR diversity of virus-specific CD8 T cells provides the mechanistic basis for massive TCR renewal after antigen exposure. J Immunol 186, 7039-7049 (2011).

19. Price, DA et al. Avidity for antigen shapes clonal dominance in CD8 + T cell populations specific for persistent DNA viruses. J Exp Med 202, 1349-1361 (2005).

20. Retiere, C. et al. Generation of cytomegalovirus-specific human T-lymphocyte clones by using autologous B-lymphoblastoid cells with stable expression of pp65 or IE1 proteins: a tool to study the fine specificity of the

Claims (7)

1. A method of predicting the status of a human leukocyte antigen (HLA) allele of a subject, comprising: (a) determining an immune receptor profile of unique T-cell receptor (TCR) rearranged DNA sequences for each of a plurality of subjects, each subject having a known HLA allele status; (b) categorizing the plurality of subjects based on (i) said subject's known HLA allele status and (ii) a presence or absence in the subject's immune receptor profile of a trait comprising a single TCR rearranged DNA sequence ; (c) determining a statistical score for the association between a set of traits and an HLA positive allele status based on (b); (d) training a machine learning model using said set of traits to define a set of classifiers for each HLA allele state; (e) introducing one or more rearranged DNA sequences from the unique TCR of a subject with unknown HLA allele status into said machine learning model to identify one or more traits that match the set of classifiers; and (f) predicting the status of an HLA allele of said subject based on said one or more paired traits. The method of claim 1, wherein training a machine learning model comprises training a logistic regression model using said set of identified characteristics and said known HLA allele states of each subject. The method of claim 1 or claim 2, further comprising determining a false discovery rate (TDF) of the association of a trait with an HLA allele status. The method of any one of claims 1 to 3, further comprising determining a series of false positive associations between said trait and said HLA allele status. The method of any one of claims 1 to 4, wherein training a machine learning model further comprises performing a cross-validation method of leaving one out. The method of any one of claims 1 to 5, wherein said HLA allele is an HLA-A2 allele or an HLA-24 allele. The method of any one of claims 1 to 6, wherein determining an immune receptor profile comprises determining the total number of unique TCR sequences and the frequency of each unique TCR sequence.